fix!(parse_json): (lossy) convert large numbers to float to avoid int…

…eger overflow (#570)

CHANGELOG.md

@@ -2,6 +2,7 @@
 
 ## unreleased
 * `parse_regex_all` `pattern` param  can now be resolved from a variable
+* fixed `parse_json` data corruption issue for numbers greater or equal to `i64::MAX` 
 
 ## `0.8.0` (2023-10-31)
 

src/stdlib/parse_json.rs

@@ -244,9 +244,8 @@ fn type_def() -> TypeDef {
 
 #[cfg(test)]
 mod tests {
-    use crate::value;
-
     use super::*;
+    use crate::value;
 
     test_function![
         parse_json => ParseJson;
@@ -310,5 +309,18 @@ mod tests {
             want: Err("max_depth value should be greater than 0 and less than 128, got 129"),
             tdef: type_def(),
         }
+
+        // // TODO: provide a function version of the `test_function!` macro.
+        max_int {
+            args: func_args![ value: format!("{{\"num\": {}}}", i64::MAX - 1)],
+            want: Ok(value!({"num": 9_223_372_036_854_775_806_i64})),
+            tdef: type_def(),
+        }
+
+        lossy_float_conversion {
+            args: func_args![ value: r#"{"num": 9223372036854775808}"#],
+            want: Ok(value!({"num": 9.223_372_036_854_776e18})),
+            tdef: type_def(),
+        }
     ];
 }

src/value/value.rs

@@ -1,10 +1,9 @@
 //! Contains the main "Value" type for Vector and VRL, as well as helper methods.
 
-use std::collections::BTreeMap;
-
 use bytes::{Bytes, BytesMut};
 use chrono::{DateTime, SecondsFormat, Utc};
 use ordered_float::NotNan;
+use std::collections::BTreeMap;
 
 pub use iter::{IterItem, ValueIter};
 

src/value/value/serde.rs

@@ -99,8 +99,24 @@ impl<'de> Deserialize<'de> for Value {
             }
 
             #[inline]
-            fn visit_u64<E>(self, value: u64) -> Result<Value, E> {
-                Ok((value as i64).into())
+            fn visit_u64<E>(self, value: u64) -> Result<Value, E>
+            where
+                E: serde::de::Error,
+            {
+                if let Ok(value) = i64::try_from(value) {
+                    Ok(value.into())
+                } else {
+                    // TODO: Address this issue by providing a lossless conversion option.
+                    #[allow(clippy::cast_precision_loss)]
+                    let converted_value = value as f64;
+                    let wrapped_value = NotNan::new(converted_value).map_err(|_| {
+                        SerdeError::invalid_value(
+                            serde::de::Unexpected::Float(converted_value),
+                            &self,
+                        )
+                    })?;
+                    Ok(Value::Float(wrapped_value))
+                }
             }
 
             #[inline]